Province-Wide Tapping of a Shallow, Variably Depleted, and Metasomatized Mantle to Generate Earliest Flood Basalt Magmas of the Columbia River Basalt, Northwestern USA

Abstract

The Miocene Columbia River Basalt Group (CRBG) of the Pacific Northwest of the United States is the world’s youngest and smallest large igneous province. Its earliest formations are the Imnaha, Steens, and now the Picture Gorge Basalt (PGB), and they were sourced from three different dike swarms exposed from SE Washington to Nevada to northcentral Oregon. PGB is often viewed to be distinct from the other formations, as its magmas are sourced from a shallow, relatively depleted, and later subduction-induced metasomatized mantle, along with its young stratigraphic position. It has long been known that the lowermost American Bar flows (AB1&2) of the Imnaha Basalt are chemically similar to those of the PGB, yet the Imnaha Basalt is generally thought to carry the strongest plume source component. These opposing aspects motivated us to revisit the compositional relationships between AB1&2 and PGB. Our findings suggest that tapping a shallow, variably depleted, and metasomatized mantle reservoir to produce earliest CRBG lavas occurred across the province, now pinpointed to ~17 Ma. Moreover, compositional provinciality exists indicating regional differences in degree of depletion and subduction overprint that is preserved by regionally distributed lavas, which in turn implies relatively local lava emplacement at this stage.

1. Introduction

The Miocene Columbia River Basalt Group (CRBG) of the Pacific Northwest of the United States is the world’s youngest and smallest large igneous province (LIP) (e.g., [1,2]) (Figure 1). The Group consists of seven formations (Figure 1 and Figure 2), formally defined by [3] and distinguished on the basis of stratigraphic position, distribution, lithology, and chemical composition [4]. Despite extensive study, numerous questions remain in the collective understanding of the Columbia River Basalts (CRBs). Such questions include: (1) the exact driving force of this magmatism, mantle plume or plate tectonics (e.g., [5,6,7,8,9]), (2) the precise timing of CRBG magmatism [2,10,11,12,13,14], (3) the nature of crustal storage sites prior to eruption [5,15,16,17]; and (4) chemical variations among lavas and what these imply about mantle sources [18,19,20,21,22,23].

Current understanding of the stratigraphy and spatio-temporal evolution of the CRBG has evolved over many decades [1,3,26,27,28,29,30], often involving significant revisions to earlier schemes. In early work [27], the name “Picture Gorge Basalt” was applied to the early flow package underlying the Grande Ronde lavas (called “Yakima Basalt” at the time) in the deep canyons of northeastern Oregon, western Idaho, and southeastern Washington, now assigned to the Imnaha Basalt [3,26]. However, the type locality for Picture Gorge Basalt (PGB) is located approximately 150 km to the west and consists of lavas originating from the Monument dike swarm rather than the Chief Joseph swarm, producing the Imnaha Basalt and overlying formations (Figure 1). Chemical and mineralogical differences between the eastern basalts and type Picture Gorge Basalt lavas were recognized and were key to establishing these lavas as their own formation, the Imnaha Basalt [3,26,29]. Nonetheless, strong similarities exist between some Imnaha flows and PGB, but whether they had the same age, came from the same source, and were physically continuous with PGB flows was uncertain [31]. At that same time, however, new radiometric age determination [32] indicated that Picture Gorge Basalt was significantly younger than the Imnaha Basalt and contemporaneous with eruption of the middle of the Grande Ronde Basalt that overlies the Imnaha Basalt (Figure 2). As a result, the most PGB-like, at the base of the whole basalt flow sequence in the region of Imnaha, were grouped with the overlying flows into the Imnaha Basalt [3,26] and this has remained the case ever since. Recent work by [2] extended the PGB formation to the east (Figure 1) and yielded ⁴⁰Ar/³⁹Ar ages that indicate that eruptive activity of the Picture Gorge Basalt started significantly earlier around 17.25 Ma, making the PGB older than or, as a minimum, as old as the earliest CRBG lavas anywhere. The updated timeframe now makes it probable that the lower flows of the Imnaha Basalt and the Picture Gorge Basalt may have erupted synchronously, as originally envisaged by [3,27]. Ref. [2] also drew attention back to these lowest two flow units of the Imnaha and questioned whether they might be far-travelled PGB flows or erupted more locally where they are exposed. This context provided the impetus for this study, in which we revisit the compositional relationships between the lowest flow units of the Imnaha Basalt and the Picture Gorge Basalt and the other Imnaha lavas that occur stratigraphically higher.

Although our focus is quite specific about lava compositions and stratigraphy of select locations, it is significant for the understanding of temporal and spatial trends of CRBG magmatism across this flood basalt province. PGB magmas have long been viewed as being sourced from a shallow depleted mantle that has been metasomatized in a subduction setting prior to the CRB event [11,19,20,33] whether or not this mantle is lithospheric is controversial [11,20]. PGB-like lavas of the Imnaha as well as the earliest Steens Basalt lavas from the southern CRBG extent (Figure 1) share compositional characteristics with PGB lavas. Hence, it appears that tapping a shallow mantle reservoir gave rise to the earliest CRBG lavas over a much broader area than previously appreciated. Furthermore, the longer eruption duration recently documented for the PGB [2] indicates longer term prevalence of this depleted mantle source, and it also appears there is compositional evidence in the stratigraphy of the Imnaha and Steens Basalt that also record (although muted and more punctuated) the tapping of a depleted yet metasomatized shallow reservoir through time.

2. Geologic Background

2.1. The Imnaha Basalt

The currently accepted stratigraphy of the Imnaha Basalt was established by [26] who subdivided the formation into two main chemical types: the American Bar (AB) and Rock Creek (RC). Within the American Bar type, nine flow units are distinguished, AB1 through AB9, while the Rock Creek type is represented by RC1 through RC3, plus the Fall Creek (FC) flow and the capping Log Creek (LC) flow. AB1 and AB2 are the stratigraphically lowest flows of the Imnaha Basalt. Higher in the composite section, AB and RC flows are intercalated, with AB dominating the lower part of the sequence and RC the upper (Figure 2) [26].

Lavas of Imnaha Basalt, especially the basal flows, are confined to the lowest parts of the greater Hells Canyon area, where they are underlain by accreted terrane rocks and overlain by flows of the Grande Ronde Basalt (Figure 2, Figure 3 and Figure 4). Ref. [26] noted that stratigraphic correlations between sections are generally good north of the Wallowa Mountain-Seven Devils divide, while less consistent south of it (Figure 4). But even in the north, there are Imnaha sections that start with different flows, and some flows can be missing. Stronger pre-Imnaha paleotopographic relief is likely the main reason for missing lower flows. Upward missing flows could also be explained by lava flowing only into specific paleo drainages.

All Imnaha lavas range from olivine to quartz-normative tholeiites with a concentration of 4.04–7.09 wt.% MgO [11]. Lavas of the Rock Creek type have a range of 48–51 wt.%, and the American Bar type have 51–52.5 wt.% SiO₂ (on 100% volatile free basis); thus, both subtypes can be readily distinguished by silica content (Figure 5) [26]. On average, American Bar lavas have lower Ni content, higher Sc and Ca, and lower abundances of incompatible trace elements than Rock Creek samples, seen most clearly in the two lowest AB flows (AB1&2). Recently, [23] have shown that the differences between the two types can be explained by variable mineralogy in the source mantle; the RC source had a higher pyroxene to olivine ratio than that for AB. The difference is most clearly seen in the Ni contents of olivine from the two types, with AB flows having low-Ni olivine. There is some overlap between the two, and one flow (Log Creek) contains olivine of both types.

Radiogenic isotope ratios of Imnaha lavas lie at the apex of radiating trends defined by all other CRBG formations, suggesting a mantle source dominated by one component. This enriched component was identified by [18], and attributed to a mantle plume source by [16,20].

2.2. The Picture Gorge Basalt

The PGB, as defined by [3] erupted from the Monument dike swarm in north-central Oregon (Figure 1). Its volume is 3300 km³ and is relatively small, accounting for only 1.1% of the CRBG as compared to the other major units like the Imnaha Basalt and Grande Ronde Basalt, which comprise 15% and 72% of the CRBG, respectively [1]. However, recent work by [24] has revealed that the extent of PGB lavas is likely larger by ~14,000 km² with a total distribution of ~25,000 km² (Figure 1), yielding an additional volume ranging from ~3000–8000 km³ based on using different thickness estimates.

The primary and original outcrop area of PGB flows is the John Day Basin of the Blue Mountains region of north-central Oregon (Figure 1) [34]; this has now been extended to the south and east to reach Malheur Gorge (Figure 1) [24], where Imnaha Basalt, Grande Ronde Basalt, and Steens Basalt interfinger (Figure 1) [30,35]. The northwestern outcrop area of the original PGB was also extended eastward based on new data [24]. Until recently, emplacement of the PGB was thought to have taken place 16.4–15.2 Ma [10]. New data now constrain PGB activity to 17.23–15.76 Ma [2,24].

Although PGB is generally geochemically similar to main-phase CRBG units in major elements, it is closest in composition to the Steens Basalt with respect to incompatible trace elemental and isotopic compositions [2,11,19,24]. Picture Gorge Basalt shows comparable SiO₂ wt.% ranges of 49–53 wt.% to those of the Steens and Imnaha Basalt, but PGB contains lower concentrations of Th, high field strength elements (HFSEs), light rare earth elements (LREEs) and Zr/Y values [2]. Compatible trace element abundances are however more similar to Imnaha AB than to Steens; this extends to the Ni content of olivines, suggesting that the PGB is derived from a peridotite source similar to AB lavas [23]. The PGB is primarily thought to be the product of a subduction-modified MORB-like source [11,16,19,20,36,37].

2.3. Other Main Phase CRBG Formations: Steens Basalt, Grande Ronde Basalt, and Wanapum Basalt

Here, for the sake of completeness, we briefly summarize these three additional main phase formations without treating each separately as they are less the topic of this paper. Based on previous geochronological work, onset of eruption of lavas of the Steens Basalt slightly predates the ones of the Imnaha Basalt at 16.97 ± 0.06 Ma [25]. All Steens Basalt lavas are thought to have erupted from the Steens dike swarm in southeastern Oregon (Figure 1). Steens Basalt is subdivided into the Upper and Lower Steens, which itself is divided into Lower Steens A and B [25,38]. Lower Steens Basalt generally reflects a relatively primitive composition, characterized by tholeiitic basalts interpreted to represent a homogenous mix of melts derived from depleted and enriched mantle with little contamination from the crust. By contrast, the Upper Steens Basalt is compositionally more alkalic and enriched in incompatible elements compared to the Lower Steens Basalt [25,37]. The Steens Basalt is thought to represent a mix of an enriched, OIB-like plume component and a depleted mantle component [11,22,38]. The compatible element inventory of Steens lavas nonetheless suggests derivation from a pyroxenite source, more similar to that for the Imnaha RC type than PGB [23].

The Grande Ronde Basalt is volumetrically the largest unit of the CRBG, covering 170,000 km² and comprising 72% of the CRBG by volume [1]. Eruptions of the Grande Ronde commencing ca. 16.56 Ma [14] from the Chief Joseph dike swarm (Figure 1). In the Hells Canyon region, the Grande Ronde Basalt conformably overlies the Imnaha Basalt. In the field, the Grande Ronde Basalt can be differentiated from the Imnaha Basalt as generally finer-grained and more resistant to erosion; additionally, most flows lack large phenocrysts. Chemically, the Grande Ronde is tholeiitic and more evolved than the Imnaha Basalt, with SiO₂ values generally above 53 wt.% and MgO values generally below 5 wt.% (Figure 5) [1,11,16]. The trace element and isotopic characteristics of the Grande Ronde show the influence of crustal contamination, and their genesis has been attributed to combined assimilation and fractional crystallization of Imnaha-type parental magmas [11,16,18].

The Wanapum Basalt consists of six members which are, in stratigraphic order: Ekler Mountain, Lookinglass, Frenchmen Springs, Shumaker Creek, Roza, and the Priest Rapids member. The Frenchman Springs, Roza, and Priest Rapids are major members, each having volumes >1000 km³ and SiO₂ contents of ca. 50%, while the more primitive Eckler Mountain and more evolved Lookinglass and Schumaker Creek members are much smaller. Earlier age constraints suggested an activity period from 15.5 to ~15 Ma cf. [1] and hence lavas of the Wanapum Basalt were considered part of the CRBG “waning phase”. More recent age dating of intercalated tuffaceous sediments however suggests that earliest Wanapum lavas erupted as early as 16.1 Ma and lavas of the youngest member around 15.9 Ma [12]. The latter age is supported by U-Pb zircon and sanidine ⁴⁰Ar/³⁹Ar ages and tephra correlation of silicic ashes overlying the capping Priest Rapids Member of ~15.8–16.0 Ma [39,40,41,42]. Given these new ages, it is reasonable to include the Wanapum Basalt in the “Main Eruptive Phase” cf. [21]. Lavas of the Wanapum Basalt cover about 87,400 km² and crop out mostly in Washington state and along the Washington/Oregon state border. Lavas represent a volume of 5.3% of the total CRBG volume [1]. Lava compositions are diverse, ranging from early Eckler Mountain basalts with ~8 wt.% MgO through the major members with typically ~4–5.5 wt.% MgO, to the minor basaltic andesite Schumaker Creek lava with >3 wt.% MgO. Trace element and isotopic variations among the Wanapum lavas suggest an origin involving complex fractionation and interaction between primitive Eckler Mountain-type magma and residues from the preceding Grande Ronde magmatic episode [43].

3. Methods

3.1. Fieldwork

Sampling for this study was carried out in two areas, the northern one encompassing the greater Hells Canyon area located north of the Wallowa–Seven Devils Mountains divide (Figure 4). The northern area has historically received the most attention for investigating Imnaha Basalt and is the basis for establishing the existing stratigraphy and chemical types of the Imnaha Basalt [26]. The second area includes a wide swath from just south of the Wallowa–Seven Devils Mts divide (Figure 4) towards the town of Huntington and the town of Brogan. Lavas of Imnaha Basalt have been reported from this area e.g. [44], and this area constitutes the southeastern-most extent of the Imnaha Basalt (Figure 1). In both areas, lavas of Imnaha Basalt directly overlie rocks of Paleozoic terranes or Cretaceous plutonic (mostly granodioritic) rocks.

Sampling locations in the northern study area were chosen based on stratigraphic columns, sample locations, and compositional data from [26]. A total of nine locations were sampled throughout the greater Hells Canyon area of NE Oregon and western Idaho (Figure 3). We resampled locations studied by [26] that previously indicated exposures of American Bar flows 1 and 2, as these flows are not as uniformly distributed throughout the Hells Canyon area compared to stratigraphically higher American Bar and Rock Creek flows. This is mainly due to the significant pre-existing relief at the time of onset of CRBG eruption in NE Oregon (Figure 3). Resampled locations of [26] include Dug Bar, Eagle Creek, China Creek, Riggins, and Whitebird Creek. Several additional transects were conducted at locations proximal to terrane rock–basalt boundaries, in the northern area these sites being Eagle Creek and Tully Creek; south of the Wallowa Mountains, additional transects were performed at Richland, Slaughterhouse Range, and Willow Creek. Lastly, one or two flows near terrane boundaries were sampled at Skookumchuck, Brownlee Dam, Big Lookout Mountain Road, and Huntington Road (Figure 4).

3.2. Analytical Work

The majority of samples collected in the field were analyzed for their major and trace element compositions by X-ray fluorescence (XRF) and inductively coupled plasma mass spectrometry (ICP-MS) analyses at the Peter Hooper GeoAnalytical Laboratory at Washington State University, Pullman. Samples were first chipped and further prepared according to laboratory standards, depending on the type of analysis. Major and trace element data were determined with a ThermoARL X-ray fluorescence spectrometer, and trace elements and REEs were analyzed with an Agilent 7700 ICP-MS. More detailed descriptions of analytical methods and precision can be found in [45] for XRF and [46] for ICP-MS, available at https://environment.wsu.edu/facilities/geoanalytical-lab/technical-notes/ (accessed 20 February 2021).

Petrographic thin sections were prepared by Spectrum Petrographic. Thin sections were examined to determine mineral phases and relative abundances.

Three samples from the southern sector of the study area were chosen for age dating. New high-precision ⁴⁰Ar/³⁹Ar ages (

Table 1. Summary Table of ⁴⁰Ar/³⁹Ar Ages for Imnaha Basalt.

Sample Location	Sample Name	Phase	Age	Error	MSWD	40Ar/36Ar	No. Steps
			(Ma)	(±2σ)		Intercept
Richland section	MM-CRB-32	plagioclase	17.11	0.15	1.60	295.11	9
Richland section	LF-21-54	groundmass	17.18 *	0.07	1.76	303.10	13
Willow Creek section	LMF-19-80	groundmass	16.84 *	0.07	4.43	296.16	7
Plateau ages are within error of their inverse isochron age unless otherwise specified

* = Mini plateau.

) were obtained by incremental heating methods using the ARGUS-VI mass spectrometer housed at the Oregon State Geochronology Lab. Samples were irradiated for 6 hours (Irradiation 15-OSU-07) in the TRIGA CLICIT nuclear reactor at Oregon State University, along with the FCT sanidine (28.201 ± 0.023 Ma, 1σ) flux monitor [47]. Detailed analytical procedures, geochronological methods, and data are included in the Supplemental Material (S4, S5).

3.3. Assignment of Imnaha Chemical Types

As our sampling targeted the stratigraphic lowest part of the Imnaha Basalt, our evaluation focused on determining if collected samples fall into the following categories: (1) they represent lavas of American Bar flow 1 and 2 (AB1&2); (2) they represent middle to higher American Bar flows (AB3+), or (3) they represent flows of the Rock Creek type (RC). In our comparison, we initially excluded data from the stratigraphically highest Rock Creek flows, the Fall Creek and Log Creek chemical types, as both of these types exhibit chemical systematics overlapping with lower Imnaha flows, yet our stratigraphic focus on the basal Imnaha flows falls clearly below where Fall or Log Creek flows are exposed. Where we sampled all the way up to the overlaying Grande Ronde Basalt flows (Richland section), we specifically address whether or not flows of these two types were sampled there. However, we will revisit the top Imnaha flow units later in the discussion.

The chemical data and flow classifications from [16,26] were used to compare the data of this study to determine the likely flow designation of samples collected. We used a combination of major and incompatible trace elements to designate the chemical affinity of our samples. Examples are shown in Figure 6 and Figure 7. Plots for all sample locations as well as additional plots used for assigning a unit identity can be found in the Supplemental Material (S1, S2, S3).

4. Results

4.1. Dug Bar

Dug Bar is the type locality for the Imnaha Basalt [26]. Here, the Snake River incised through around a kilometer of basalt flows to the current water level. On the northern and southern margins of Dug Bar, the Imnaha Basalt is underlain by accreted terrane rocks comprised of granodiorites, metasedimentary, and metavolcanic rocks of the Wallowa terrane. Directly across the river from the sample location, the basalts form a series of colonnades and entablatures indicative of differential cooling directions [49].

The seven samples, likely representing multiple flows (up to seven) collected at Dug Bar are porphyritic, containing elongate plagioclase phenocrysts ranging from 5 mm to 20 mm in length within a fine plagioclase groundmass with pyroxene and minor olivine and pyroxene phases [49] (S1 Supplemental Material).

All seven samples were selected for bulk rock analysis. These samples have SiO₂ values that range from 51.2 to 51.9. The five samples from the stratigraphically lowest flows are nearly indistinguishable, while the two uppermost are compositionally distinct (Figure 6, S2, S3 Supplemental Material). For example, the elemental range of these five for TiO₂, CaO, La, and Zr are, respectively: 1.68–1.72 wt.%, 10.36–10.58 wt.%, 9.9–10.2 ppm, and 118 to 121 ppm. The values for the same elements for the stratigraphically highest two samples are: 1.96–2.02 wt.%, 9.68–9.79 wt.%, 15.2–16.8 ppm, and 168–172 ppm. All samples possess Mg numbers of 46, with the exception of the two stratigraphically highest samples with values of 40 and 43. The five lower samples all fall within the envelope representing the spread of samples classified as American Bar 1 and 2 by [26], while the higher pair plot in diagnostic major element space (TiO₂ vs. SiO₂ or TiO₂ vs. CaO) into the American Bar 3+ field (Figure 6). The silica values for all of the Dug Bar samples reported are too high to represent Rock Creek flows [26].

The trends of samples from Dug Bar seen on normalized incompatible element diagrams (‘spidergrams’) are consistent with the grouping based on major elements (Figure 7). The majority of the samples plot more closely to the AB1&2 average values. Samples LF-20-05 and LF-21-40 (the “two higher” units), while not quite as enriched, are most similar to the AB3+ pattern. This is most noticeable with the negative Sr and Ti anomaly observable in the average AB3+ curve that is not present in RC flows, while a positive Sr spike appears in the first two AB flows.

All samples have peaks in Ba and Pb, with sample LF-20-05 being enriched in Pb by nearly an order of magnitude. Additionally, all samples from Dug Bar have a relatively distinct Nb-Ta trough that is present throughout the Imnaha Basalt.

4.2. Tully Creek

Five samples were collected along the west bank of the Imnaha River proximal to Tully Creek, located approximately 10 km south-southwest of Dug Bar (Figure 4), and all are near the same stratigraphic level (S1 Supplemental Material). Like the exposures at Dug Bar, the basalt outcrops at Tully Creek form entablatures underlain and capped by colonnades. The basalts are again porphyritic, with elongate 5 to 25 mm twinned plagioclase phenocrysts in a fine plagioclase-pyroxene matrix with minor olivine, pyroxene, and opaque phases [49].

The samples from Tully Creek all plot tightly together (e.g., with SiO₂ values of 51.39 to 51.51 and TiO₂ values of 1.56–1.69 wt.%). Observed concentrations are consistent with values for samples of American Bar 1 and 2 (S2 Supplemental Material). The incompatible element patterns for the Tully Creek samples also most closely resemble those of the lowest American Bar flows 1 and 2 (S3 Supplemental Material). The only readily discernible differences between the average AB1&2 values and the values of Tully Creek samples are the marginally elevated concentrations in the most incompatible elements and a small positive K anomaly likely reflecting weathering effects.

4.3. Eagle Creek

Eagle Creek is the northernmost location visited during our sampling campaign. Three samples were collected at this location along a creek bed that drains into the Salmon River (S1 Supplemental Material). Outcrops are not as pronounced along Eagle Creek, with many being eroded and buried by weathered basalt colluvium, yet samples were taken from progressively higher stratigraphic levels. Samples collected at Eagle Creek once again exhibit large plagioclase phenocrysts typical of the lower basalt flows throughout Hells Canyon [49].

Major element values for these samples are more variable than reported for the previous locations. Silica values range from 50.4 wt.% to 51.3 wt.%, and TiO₂ ranges from two values around 1.7 to one with 2.1 wt.%. Lower TiO₂ but higher CaO in two samples compared to the third sample correspond well with incompatible element concentrations (S2, S3 Supplemental Material), suggesting that these two samples correlate with AB1&2 compositions while the third sample reflects a AB3+ composition rather than a Rock Creek flow.

Sample LF-21-29 is somewhat anomalous, in that is has a relatively low Rb value compared to the average values for the other Imnaha Basalt flows. The sample also possesses a negative K anomaly, indicating that this is likely the result of post-eruption alteration rather than representative of the original composition.

4.4. China Creek

China Creek is located roughly 3.5 km to the west of Eagle Creek. At this location we were able to collect samples across a stratigraphic transect of flows proximal to the Salmon River. We collected eight samples sequentially from the lowest exposed basalt (LF-91-27) towards progressively higher stratigraphic levels (S1 Supplemental Material). As at Eagle Creek, basalt exposures here are partially covered by colluvium and weathered basalt. Basalts are all porphyritic, with large 1 to 2 cm elongate plagioclase phenocrysts in a fine groundmass [49].

In major element space of TiO₂ vs. CaO and TiO₂ vs. SiO₂, it is evident that five samples fall into or very close to the AB1&2 field, as they have low TiO₂ and high CaO and SiO₂, comparable to AB1&2 samples. Three samples are clearly different and fall into the AB3+ field and Rock Creek field in the TiO₂ vs. CaO plot. However, the TiO₂ vs. SiO₂ identifies them clearly as AB3+ samples (S2 Supplemental Material).

Incompatible element compositions of the lowest five China Creek samples are consistent with findings based on major elements, as their elemental pattern plots on top of the average AB1&2 curve. The remaining three uppermost samples closely match the patterns for Rock Creek and AB3+, yet the higher silica values suggest that they belong to the latter (S3 Supplemental Material).

4.5. Riggins and Eagle Gulch

The Riggins sample, LF-21-03, was collected from the colonnade near the town of Riggins [49]. Samples LF-21-01 and LF-21-02 were collected to the north at Eagle Gulch.

The major element plots show the Riggins sample plotting close to AB1&2. The other two samples are more ambiguous, with silica concentrations midway between average Rock Creek and AB averages and Mg numbers closest to AB3+ (S2 Supplemental Material). The low CaO and high TiO₂ values seen in these samples are closest to Rock Creek samples.

The incompatible element pattern for the sample from Riggins is most similar to the AB1&2 average (S3 Supplemental Material). The other two samples from Eagle Gulch are at the high end of the sigma 1+ envelope of avg. AB3+ and Rock Creek samples. The steeper elemental patterns from P through Lu suggest they are rather Rock Creek compositions, but the Sr trough resembles AB3+. Overall higher incompatible elements and lower Sr would be however consistent if they represent slightly more fractionated Rock Creek compositions. Mg# values at the lower end of Rock Creek composition would also be compatible with this. Taken all together, we group these two samples with the Rock Creek unit.

4.6. Pittsburg Saddle

The Pittsburg Saddle sampling location site is unique in that it is stratigraphically removed from other nearby sampling locations by thrust faulting that occurred directly to the east of the sample site, uplifting these exposures [50]. All samples collected at this location plot most closely to the AB1&2 compositions, with average values around 51.5 wt.% silica and low TiO₂ concentrations. CaO and TiO₂ values cluster closely together, around 10.3 wt.% CaO and 1.67 wt.% TiO₂ (S1, S2 Supplemental Material).

The incompatible element patterns for the Pittsburg Saddle samples plot close to one another and along the upper sigma 1+ envelope to the average AB1&2 values. Hence, major and trace elements are consistent to group them with AB1&2 (S3 Supplemental Material).

4.7. Skookumchuck Creek

Three samples were collected from a cliff face proximal to Skookumchuck Creek, with the lower two samples likely sampling bottom and top of a lower flow and the third sampling the overlying flow [49]. The basalts at this location also form colonnades and contain large plagioclase grains in a fine-grained plagioclase matrix.

These samples exhibit major element data which are most similar to the average values for the Rock Creek flows, ranging from 49.5 to 50.5 wt.% silica. Major elements CaO and TiO₂ values display a similar distribution with samples plotting near the Rock Creek and AB3+ average values, with CaO values ranging from 8.7 wt.% to 9.7 wt.% and TiO₂ values of 2.8 wt.% (S2 Supplemental Material).

Considering the low silica content and incompatible element patterns, these samples are most similar to Rock Creek. Samples LF-21-04, and LF-21-06 to a lesser extent, have substantial Rb depletions. Both samples have a negative K anomaly as well, likely indicators of secondary alteration, as Rb and K are both fluid-mobile elements (S3 Supplemental Material).

4.8. Whitebird Creek

Five samples were collected along a short transect south of the town of Whitebird (S1 Supplemental Material). Two samples, LF-21-14 and LF-21-15, most closely resemble the average values for AB3+, while the remaining samples, LF-21-16, 18, and 19, are most similar to Rock Creek (S2 Supplemental Material). Silica values are distributed in two groups; the three samples taken higher in stratigraphy contain around 49–50 wt.% silica, while the lower two samples contain 51.5 wt.% silica. A similar pattern can be seen with TiO_2, where the stratigraphically lowest samples contain lower TiO₂, around 2.1 wt.%, while the three higher samples all contain just over 3.0 wt.%. The lowest samples also contain slightly higher CaO than the samples from higher flows.

All samples collected from Whitebird are most similar to the normalization patterns representing the average AB3+ and Rock Creek flows (S3 Supplemental Material). The stratigraphically lower samples, LF-21-14 and 14, both have incompatible element values slightly below those of the average values of AB3+, while the stratigraphically higher samples are all slightly enriched compared to the average values of Rock Creek. Samples LF-21-16, 18, and 19 may have experienced some loss of K due to slight weathering, leading to lower than expected normalized K values in element patterns of these samples. All samples collected at this location possess negative Sr anomalies. With that in mind, samples LF-21-16, 18, and 19 resemble Rock Creek samples from Riggins, where we suggest negative Sr anomalies with plotting along upper envelope of Rock Creek indicates more fractionated Rock Creek compositions. The Whitebird Rock Creek samples also indicate lower Mg#.

4.9. Richland

Ten samples were collected along a transect to the southeast of Richland, Oregon (S1 Supplemental Material). One additional sample was taken ~300 m up the road from the transect location. The Richland location is roughly 130 km south from the sites sampled to the north, and is the second northernmost location of our second sampling area located south of the Wallowa–Seven Devils Mountains divide (Figure 4). The basal CRBG flow here directly overlies Burnt River schist that crops out ~50 m down the road from the transect location and that represents accreted terrane rocks. The lower flows of this outcrop are partially covered in colluvium. The basalts are slightly vesicular and contain cm-scale plagioclase phenocrysts similar to the basalts found at the sample sites to the north. About midway in the section, some parts are more friable as they are more vesiculated and some are brecciated, representing flow tops or bottoms. The uppermost flows, near the top two samples, are finer-grained, dense and are phenocryst-poor to aphyric. Lithological variations along this transect by itself suggest that the lower two-thirds along the transect are Imnaha flows while the uppermost part of the transect and the overlying towering cliffs are lavas of the Grande Ronde Basalt [49].

The samples collected along this transect show an array of compositions, some unique compared to those documented for the sample locations of the northern area. Seven samples possess CaO, TiO₂ (with the exception of LF-21-54; (S2, S6 Supplemental Material)) and incompatible element patterns very similar to AB1&2, yet have SiO₂ and Mg# that resemble those of Rock Creek lavas, except LF-21-50 that has a SiO₂ content of 50.8 wt.%, matching AB1&2 values (S6 Supplemental Material). LF-21-51, -52, are most similar to Grande Ronde Basalt, in line with their aphyric texture. Sample LP-21-53 is texturally like these two and is our stratigraphically highest sample. It has a SiO₂ content of 52.2 wt.%, which is not typical for Grande Ronde Basalt. The remaining samples appear most similar to the AB3+ subset.

Incompatible element patterns reveal that the stratigraphically lowest five samples, the seventh sample (LF-21-50), and the sample collected further up the road (MM-CRB-32) plot closer to the average AB1&2 value, with two samples, LF-21-47 and MM-CRB-32, plotting substantially below. All these samples have a Nb-Ta trough and have also notably stronger Ba and Sr peaks (S3 Supplemental Material). While we group these samples as AB1&2, we recognize that there are distinct differences to our AB1&2 samples of our northern area that will be addressed below.

4.10. Slaughterhouse Range

A subset of eight samples were selected for analysis from a sample transect about 8 km east of Huntington, OR (S1 Supplemental Material). Accreted terrane rocks also underlie this section. Basalts are all porphyritic with ~10% phenocrysts of plagioclase and a coarse-grained groundmass, except for the most SiO₂-rich sample, LMF-19-73, which is fine-grained with only small and few phenocrysts (<5%), and the highest sample LM-19-76 which has also few phenocrysts [51]. Similar to the Richland samples, compositions observed here are diverse and somewhat deviate from what was observed in the northern sampling area. The two stratigraphically lowest samples (LMF-19-70, -71) and the fifth sample LMF-19-74 are overall best matched with lower American Bar (i.e., AB1&2) compositions, yet significant differences exist (S2 Supplemental Material). The strongest evidence for this association is seen in mantle-normalized incompatible element patterns (S3 Supplemental Material). All three samples plot on top of AB1&2 average or close to the +1 sigma envelope. In terms of major elements, low TiO₂ and higher Mg# on plots vs. SiO₂ similarly suggest that, although the compositional field is scattered around the actual AB1&2 field. The other samples overall fit best with AB3+ compositions. Among these, two samples are notable; the first one is LMF-19-73 with a 53.3 wt.% SiO₂, 8 wt.% CaO, and a Mg# of only 32. Typically, such evolved compositions along with elevated incompatible elements are only observed in Grande Ronde Basalt and lavas of the upper Steens Basalt cf. [24], yet this sample is intercalated among flows with AB1&2 as well as AB3+ characteristics. Also notable is the sample from the stratigraphically highest lava, LMF-19-76, which has a trace element pattern akin to Rock Creek samples but is too high in silica for such association, unless it is again a fractionated Rock Creek composition as suggested by the Sr trough.

4.11. Willow Creek

An area about 8 km north-northwest of the town of Brogan, along Willow Creek, was investigated and samples taken (Figure 4). Here, pre-Cenozoic schists of the accreted terranes crop out along the valley floor and younger basalt forms a prominent rimrock along the western valley side as well as along the eastern side but only south of the investigated area. CRBG aged lavas crop out in between and crosscut schists in select areas. Prominent columnar jointing with variable orientations is also evident [51] (S1 Supplemental Material). All field observations together suggest that this area represents a near-surface vent area and subaerial flows. All Willow Creek samples are phenocryst-poor (<4%), with a coarse to fine-grained groundmass. Observed compositions are more restricted, although sampling around the area with columnar jointing reveals compositions that are more fractionated, also reaching 53.3 wt.% SiO₂. Trace elemental patterns are again most revealing and suggest all samples reflect AB1&2 compositions with similar differences as noted for the last two sampling sites. Major elements (TiO₂ vs SiO₂ and Mg# vs. SiO₂) of the Willow Creek samples also point to AB1&2 (or similar) compositions (S2, S3 Supplemental Material).

4.12. Brownlee Dam and Single Basal Samples from Elsewhere

Two samples were collected along the Snake River just north of Brownlee Dam that were the closest to the terrane/CRBG contact (Figure 4). This location is also the closest to northern area sampling sites. Additional samples include one sample from a basal flow on the south flank of Big Lookout Mountain, and one sample overlying terrane rocks west of Huntington (Figure 4). One of the Brownlee Dam samples is clearly AB1&2, while the second shares characteristics with AB1&2 but also has concentrations suggesting it is transitional to AB3+ but on the lower incompatible trace elements side. Major element plots suggest both are AB1&2 (S2 Supplemental Material). Similarly, both of the other two samples are best associated with AB1&2, although they have incompatible patterns as documented for the other southern areas. Interestingly, the Brownlee Dam samples are more similar to the original pattern of AB1&2 (S3 Supplemental Material).

4.13. Summary of Unit Classification

In Figure 8, we compare our samples from all sample locations and their unit assignment with the data from [26] that were reanalyzed by [16] (see also S7 Supplemental Material). This comparison indicates the following. All lavas from the north area assigned as AB1&2 closely match AB1&2 samples reported by [16] and the average given in [26], with the exception of one Eagle Creek sample that displays lower SiO₂ wt.%. Lavas from the south area assigned as AB1&2 show a much greater compositional range but incompatible element concentrations (e.g. Nb, Th, La, Zr), and the concentrations of the characteristic major element TiO₂ are consistent with values reported for AB1&2 samples by previous workers. Our sample assignment to Imnaha subunits AB3+ and RC is consistent with published data. It is notable that both of these groups are only effectively separated when using silica on bivariate plots. This particularly applies to the published data, as our AB3+ and RC samples indicate generally narrower ranges which may simply be due to the fact that we did not systematically sample these subunits. Overall, however, our classification approach seems adequate to separate lower American Bar lavas, AB1&2, from middle and upper American Bar as well as Rock Creek samples. This fact is particularly well illustrated when we compare average incompatible trace element systematics. Figure 9 shows mantle-normalized incompatible trace element patterns that clearly reveal a striking contrast between AB1&2 averages for northern and southern sampling area with averages for AB3+ and Rock Creek samples across the entire area. This also shows the differences in AB1&2 compositions between the northern and southern sampling area. Lastly, it also shows close compositional affinity of both AB1&2 averages with the average of PGB samples.

4.14. New Age Dates

We selected three samples for dating via ⁴⁰Ar/³⁹Ar geochronology from the southern area (Table 1). Two samples were collected from the Richland section and resulting ages are supported by field relationships and their relative stratigraphic position. The stratigraphically lower sample (sample LF-21-54) was collected from a lava flow overlying terrane rocks at the base of the section and yields a mini plateau age of 17.18 ± 0.07 Ma, within error of the inverse isochron age (Figure 10; S5.1 Supplemental Material). The other dated Richland sample is from a lava flow stratigraphically higher in the section (sample MM-CRB-32) and yields a mini plateau age of 17.11 ± 0.15 Ma, also within error of the inverse isochron age (Figure 10; S5.2 Supplemental Material). Sample MM-CRB-32 was collected from a lava flow that corresponds to approximately the fourth sample of transect, LF-21-47 (S1 Supplemental Material); both samples exhibit a distinctly more primitive composition and as such could be correlated (Figure 9; S6 Supplemental Material). The third sample selected for geochronology was collected from the Willow Creek section (sample LMF-19-80). The Willow Creek sample yields a plateau age of 16.84 ± 0.07 Ma that is within error of the inverse isochron age (Figure 10; S5.3 Supplemental Material).

Our new ages are older than the upper age bracket for Imnaha Basalt and are overall consistent with published data. The younger age bracket for Imnaha is 16.57 Ma, determined by two U-Pb zircon ages from a tuffaceous sedimentary unit stratigraphically above Imnaha Basalt AB1&2 and supported by a ⁴⁰Ar/³⁹Ar age from a Imnaha Basalt lava flow immediately below the Imnaha/GRB contact and likely emplaced at the end of Imnaha volcanic activity [12,14]. Earlier Imnaha activity is further constrained by another U-Pb age of 16.601 Ma from a lapilli tuff intercalated between two Imnaha lava flows. Ref. [52] report a ⁴⁰Ar/³⁹Ar age of 16.85 ± 0.21 Ma for a Imnaha lava flow at the top of Squaw Butte in west-central Idaho, located approximately 100 km southwest from our Willow Creek location. While the upper Imnaha Basalt is well-dated, none of these existing ages are derived from samples representing the base of the Imnaha Basalt.

5. Discussion

We structure our discussion by reviewing compositional similarities of lower Imnaha Basalt flows with Picture Gorge Basalt, examining regional differences in widespread onset of depleted CRBG volcanism as represented by PGB and lower American Bar lavas, evaluating the influence of this mantle component through time with the implications for magma diversity within the Imnaha Basalt. We finish the discussion by comparing samples of this study to basaltic samples of the nearby oceanic Siletzia large igneous province and to typical calc-alkaline lava compositions of the Cascade volcanic arc.

5.1. Comparison of Lower American Bar Flows (AB1&2) to the Picture Gorge Basalt

The first two American Bar flow units of the Imnaha Basalt are remarkably similar to the Picture Gorge Basalt; this was the source of early speculation about their equivalence [27,31], but was recognized in subsequent work up to the present [2,16,24,31]. In fact, [2] included samples by [49] from the Brogan area to argue for PGB lavas cropping out far southeast of where PGB was thought to occur.

On most element covariation diagrams and element ratio plots, the PGB and AB1&2 values plot with PGB or in an area of overlap (Figure 11 and Figure 12). In mantle-normalized incompatible element diagrams, the pattern of average PGB only deviates from the one of our sample averages as calculated for the northern and southern area in a few specific features (Figure 9). More specifically, PGB samples possess a slight Zr-Hf trough that is not observed in AB1&2, while those lavas, in turn, have a negative P anomaly that is not observed in the average PGB. PGB lavas tend to have a steeper Th–U transition and are more enriched in Ba and K relative to AB1&2 lavas of our northern area, yet this is not the case for southern samples of this study (Figure 9). These differences are not only observed when average values are plotted but noted differences also show up in plots with Hf/Sm, Nd/P₂O₅, Th/U, Ba/Th, La/K₂O capturing the Zr-Hf trough, P anomaly, and Ba and K enrichment, respectively (S7 Supplemental Material).

Major element concentrations are less useful to differentiate PGB and AB1&2 lavas from other Imnaha subunits, although one consistent characteristic is that PGB and AB1&2 have lower TiO₂ values than either Rock Creek or middle to upper American Bar lavas. Although AB1&2 lavas have a similar range of SiO₂ values, they are separated from the AB3+ flows by generally higher MgO values, making them more similar to the higher-silica, lower-MgO Picture Gorge lavas (Figure 11). AB1&2 are higher in CaO than AB3+ and RC lavas, as is PGB (Figure 8).

Using concentrations of 27 trace elements obtained via ICP-MS, a principal component analysis (PCA) was conducted on AB1&2 samples, AB3+ (subdivided into flows 3 to 5, and 6 to 9), RC flows, and Picture Gorge Basalt lavas. We find that, statistically, AB1&2 flows are more similar to the Picture Gorge Basalt than they are to the rest of the American Bar and Rock Creek subunits of the Imnaha Basalt (Figure 13). The principal components displayed in Figure 13 account for 88% of the variance between variables (76% for PC1 and 12% for PC2). Imnaha AB1&2 is subdivided into “north” and “south” according to our sampling area. This spatial dependence on composition will be elaborated on in the next section.

To summarize, the many important compositional commonalities of AB1&2 and PGB lavas documented here tie these units together as the earliest lavas of the CRBG north of 43.5° N, and in turn indicates the first basalts to erupt were those derived from a relatively depleted mantle, yet these lavas indicate some regional variability. Similar signatures are also observed among lavas of the Steens Basalt that erupted in the southern portion of the CRBG. In fact, the first two lavas of the newly recognized lowest-most Steens Basalt unit (Lower Steens A, [25]) have an extremely similar incompatible normalization elemental pattern as AB1&2 and PGB (Figure 14A), yet again with subtle differences that on one hand makes them more similar to AB1&2 (e.g., lower Ba enrichment), others are more like PGB (Zr-Hf trough, overall lower Th, U enrichment) and others that appear unique to Steens (steeper REE pattern) (Figure 14A).

Previously, due to its distinctly high Ba/Nb and apparent contemporaneity with Grande Ronde Basalt (Figure 2), the PGB was considered distinctive among all the main phase CRBG formations, for example attributed to an episode of back-arc magma generation of uncertain relation to the rest of the CRBG by [16]. However, the newer ⁴⁰Ar/³⁹Ar ages of [2] indicate that the PGB, including lavas of the type locality, is significantly older and contemporaneous with the Imnaha Basalt. Recently, [13] has re-interpreted the age significance of the paleomagnetic relationships to suggest that a magnetic reversal near the top of the PGB sequence corresponds to that at the Imnaha—Grande Ronde boundary, rendering the bulk of the PGB, at the youngest, age-equivalent to the Imnaha. This now gives a more consistent overall picture for the whole main-phase CRBG of early relatively primitive basalts (PGB, Steens, Imnaha) giving way to more evolved compositions with time (Grande Ronde, Wanapum). Therefore, it is important to recognize now that the earliest CRBG lavas across the province show depleted signatures, yet with regionally distinct differences, but all indicating the tapping of a depleted source. We address the regional variability of this depleted mantle source in the next section.

5.2. Compositional Provinciality of Depleted and Metasomatized Mantle Signal

Both major and trace element compositions of basal Imnaha Basalt lavas of this study clearly reveal that their compositions vary regionally. Most basal Imnaha samples collected south of the 45th parallel are characterized by a greater positive Ba anomaly, possess a deeper Nb-Ta trough, and possess a slight Zr-Hf trough, characteristics of PGB (Figure 9). Additionally, they possess lower silica and higher MgO than northern AB1&2 flows (Figure 11). The few southern flows of this study that are more like the north are samples from Brownlee Dam, Big Lookout, and the basal Slaughterhouse lava. In general, however, basal Imnaha flows show a range of compositions with respect to select chemical parameters. This was also observed by [20] who collectively called these lavas “south-of-OWL basalts” (OWL = Olympic Wallowa Lineament) and noted their resemblance to PGB. Many workers [16,18,19,20,37] have noted that PGB lavas are incompatible element-depleted compositions and have argued they are derived from a depleted mantle that experienced reenrichment reflecting a subduction-related overprint. We posit that all basal CRBG flows record an upper depleted and reenriched lithospheric mantle, but there are regional differences in the degree of the depletion and of this subduction-related metasomatic overprint. In other words, earliest CRBG lavas tapped a variably depleted and metasomatized mantle. What is less agreed upon is whether this depleted mantle is asthenospheric mantle with a nearly contemporaneous subduction overprint [11,16] or if this mantle is lithospheric and the subduction overprint is much older [20,54] As mentioned above, the differences in the degree of metasomatic overprint, and the apparent age range, led to PGB being regarded as having an uncertain relationship to other main stage CRGB formations [11]. Here we argue the opposite, that the depleted characteristic of all basal lavas (e.g., Figure 14 and Figure 15) suggest a rather common upper mantle source and that the differences in degree of depletion and metasomatic overprint of this source may be related to proximity to subduction-related fluids and melts, as well as to the possible involvement of sublithospheric mantle related to the accreted terranes. Figure 16 illustrates these regional differences among AB1&2, PGB, and Steens Basalt lavas. In general, Zr/Y, La/Yb, low Nb content (as are, e.g., Rb, Th, U, LREE contents) are likely parameters that predominatly reflect the degree of the depletion of this upper mantle source (Figure 15). On the other hand, Ba/Nb (and Ba/Th, Ba/La, or Sr/P) signify metasomatic overprint, yet more evolved compositions with high ratios may also result from crustal processes during which invidual lavas could have been affected by stronger assimilation effects or other non-source changes modifying the original LILE signals (e.g., [18,22,34]. The La/Yb ratio has been used as a measure for changes in the degree of partial melting, while an increase in Tb/Yb was taken as being due to an increase in residual garnet [23]. There is clear variation in both of these parameters among main phase CRBG formations (Figure 15), suggesting lower degrees of partial melting with significant residual garnet for Steens and Imnaha Rock Creek lavas, while PGB and AB1&2 would represent higher degrees of melting with little to no garnet; other AB lavas lie in between these extremes [23]. Differences in partial melting are nonetheless unlikely to affect LILE/HFSE ratios with elements of similar incompatibility (e.g., Ba/Nb). However, partial melting degrees must still be somewhat comparable for all discussed depleted CRBG lavas in order to produce silica-saturated tholeiitic melts from rather shallow depths [55,56]. Lastly, calling upon mantle mineralogy, as demonstrated with the high- and low-Ni trend of [23], would be insufficient as sole explanation for regional compositional differences (e.g., Ba/Na, La/Yb, Zr/Y) among discussed depleted CRBG lavas, because such parameters would not be strongly affected by changes in the clinopyroxene to olivine ratio in the mantle at comparable partial melting degrees.

Lavas can clearly travel long distances, allowing a mixed record in any one section, as superposed lavas may originate from different vent locations. Incompatible element ratios nonetheless vary somewhat regionally for lavas with a generally depleted signal as discussed above (Figure 9, Figure 11, Figure 12, Figure 15, Figure 16 and Figure 17), which is a strong suggestion that regionally variably depleted and metasomatized mantle sources are surprisingly preserved in regional lava compositions. In fact, each of the general areas (PGB, Steens, AB1&2 north, AB1&2 south) seems to record its own unique combination of chemical characteristics (Figure 16 and Figure 17).

In addition to our location data, one other specific example of how flows likely reflect local variations as well as flows that travelled far distances comes from the eastern Malheur Gorge area (Figure 1). Malheur Gorge was previously highlighted as an area where Steens lavas from the south were interfingering with Imnaha and Grande Ronde lavas derived from the north [30,35]. Refs. [2,24] expanded on this and argued that PGB lavas were interfingering as well. Select Imnaha lavas from the eastern Malheur Gorge around Namorf [17] and around Gold Creek indicate the following. Some of these have compositions like AB1&2 flows of our southern area, while there are compositions (best correlated with AB3+) that have compositional signals like northern Imnaha flows (e.g., low Ba/Nb at high Nb) (Figure 15; S8 Supplemental Material). One outcome of this compositional provinciality is that Imnaha AB1&2 flows did NOT travel from north to south, at least not southward from Brownlee Dam. One venting area for those lava flows could be the Brogan—Lookout Mountain region (cf. Figure 4); vents for American Bar-type phreatomagmatic tuffs and lavas have been described from Lookout Mountain [57] (Figure 1 and Figure 4). On the other hand, upper Imnaha lava flows may have travelled southward from the north into our southern area of investigation, also reaching into the eastern Malheur Gorge area.

5.3. Magma Types of Imnaha Basalt and Relationship of American Bar Flows 1 and 2 to the Overlying Imnaha Basalt Flows

As documented above, elemental concentration ranges and some element ratios clearly discriminate AB1&2 units from the other Imnaha subunits. On the other hand, overall incompatible elemental patterns are similar to other AB units and even to units of Rock Creek (Figure 14). This raises the question: what genetic relationship exists between AB flows 1 and 2 and the upper AB flows and RC flows? Ref. [58] argued that upper American Bar compositions could be largely derived by fractionation combined with recharge from lower American Bar compositions. Furthermore, ref. [20] argued for one Imnaha source but producing melts at three different degrees of partial melting, subsequently undergoing fractionation to give rise to AB1&2, AB3+, and to RC compositions. They argued that isotopic variations among Imnaha subunits were too small to call upon different mantle sources. To the contrary, ref. [11] did see evidence for source variation based on isotopic ratios (Figure 17). Radiogenic isotope ratios of AB1&2 plot between PGB and AB3+ − RC flows (Figure 17). The range of incompatible element concentrations at constant MgO is unlikely to have been produced by fractional crystallization alone (Figure 9 and Figure 12). Some of the higher AB flows with compositions close to AB1&2 may be related by fractional crystallization to AB1&2. However, ref. [23] showed that varying mantle source mineralogy (or fractionation of primary melts at mantle depths) is responsible for some of the variation within, and overlap between, both AB and RC chemical types. This suggests that, despite similarities of AB1&2 with AB3+ and RC, AB1&2 has its unique source (or reflects a unique proportion of sources). This is consistent with all available data and argues that AB1&2 flows should not be lumped into a single source model for the Imnaha Basalt.

5.4. Recurrence of Depleted Magma Types

Our new ages of AB1&2 and the updated eruptive timeline from [2] demonstrate that PGB eruptions temporally overlaped with eruption of the Imnaha Basalt, suggesting that a relatively depleted source was tapped for the very intital Imnaha eruptions before transitioning to eruption of more typical, more enriched Imnaha basalts (AB3+) shortly thereafter. However, there are stratigraphically higher subunits of the Imnaha Basalt that are more depleted than underlying units and hence indicate a greater role of a depleted source. These are: AB4 (#BUK5 and #DB7) [16], RC2 (#W45), RC2y (#GRJ1), and RC Log Creek (several samples) from lower to higher in the stratigraphic section (Figure 14b) [16]. In summary, the Imnaha Basalt stratigraphy suggest a depleted mantle was tapped at the very beginning of Imnaha Basalt volcanism, but it also played an increased role intermittently during subsequent eruptions. An intermittent depleted mantle signal is also preserved in the geochemical signals of the stratigraphy of the Steens Basalt lava flows [25].

5.5. Compositional Context of Earliest CRBG Lavas to Siletzia LIP and Cascade Volcanic Arc Lavas

This compositional comparison provides additional evidence for tapping a shallow and variably metasomatized mantle source without clear compositional plume signal. Numerous studies have suggested that the mostly mid-Miocene flood basalt lavas of the CRBG are a result of a rising deep-sourced mantle plume impinging on the base of the North American lithosphere [59,60,61,62,63,64,65,66,67]. On the other hand, there are a number of non-plume models that ascribe flood basalt volcanism to regional upper plate tectonics [6,65] or processes associated with the downgoing slab of the Cascadia subduction zone, such as a growing tear in the subducted oceanic lithosphere [7]. Most recently, a model is gaining popularity in which CRBG flood basalts are not the expression of the initial impingement of a deep-seated rising mantle plume but a secondary upwelling with the original plume impingement dating back to the Eocene along the North American Pacific coastal area, producing a nearby oceanic plateau that is known as the Siletzia large igneous province [68]. In this model, the initial plume—lithosphere interactions began at ~53 Ma and continued to present day as the North American plate has migrated over the plume tail e.g., [8,68]. The Siletzia LIP is estimated to have a volume of 1.7–2.6 × 10⁶ km³, 8–12 times the lava volume of the CRBG and hence is more on par with volumes of other flood basalt provinces than the CRBG [68]. Siletzia basalts crop out in various places from southern British Columbia/northern Washington State to southern Oregon, and the province is imaged in the subsurface by a strong aeromagnetic high along the entire length [68]. The model is attractive, as it explains two LIPs formed side-by-side that are ~36 myr apart.

In this context, it is interesting to compare lava compositions of Siletzia basalts with those of the CRBG, particularly the ones that are the focus of this study, the earliest CRBG lavas. Doing so reveals the following. Siletzia basalts range from MORB-like, low-Ti compositions to more enriched and high-Ti compositions (Figure 18) [69,70,71,72]. And this range is thought to reflect the interaction of a depleted source with isotopic and trace element characteristics expected for a MORB source of a spreading center with a plume with a HIMU signal [70]. While low-Ti Siletzia samples share some compositional similarities with samples of this study, as all record a more depleted source, they are distinctly different with regards to other important chemical characteristics as observed in mantle-normalized incompatible element diagrams and radiogenic isotopes. All low-Ti Siletzia basalts lack the Nb-Ta and Ti trough that is so characteristic for PGB, AB1&2 north and south, and Steens (Figure 18A) (cf. Figure 5 of [71]; Figure 7 in [72]. They also typically lack the Zr-Hf trough as compared to samples of our study, although this is less consistent. Higher-Ti Siletzia basalts strongly display a HFSE enrichment with a clear OIB signature, and hence the range of Siletzia basalts fall into the MORB-OIB array in a Ba/Nb versus Nb/Zr plot (Figure 18C). Such an array is not observed in CRBG lavas, although a minute shift to higher Nb/Zr signal is observed in Rock Creek, upper AB, and some PGB samples. Siletzia samples also contrast with trends of CRBG samples in isotopic plots with ⁸⁷Sr/⁸⁶Sr (or ¹⁴³Nd/¹⁴⁴Nd) vs ²⁰⁶Pb/²⁰⁴Pb, where Siletzia samples make for a horizontal array from lower-Ti, more MORB-like samples to higher-Ti samples (Figure 18D), while CRBG samples trend mostly diagonally from samples near the C1 component of [19] towards the C2 and Imnaha (IC) components of [19] and [16], respectively. The C1 component was influenced by an incompatible element-depleted source, while C2 and Imnaha components are influenced by an enriched reservoir with a debatable source of enrichment [16,19]. On the other hand, comparing samples of this study to typical calc-alkaline lavas of the Cascade volcanic range reveals great similarity in incompatible element patterns, such as LILE enrichment and HFSE depletion, and isotopic composition (Figure 18B–D).

The discussion above reiterates the following aspects of PGB, AB1&2 north and south, and discussed Steens lavas. Low HFSE and radiogenic isotopes trending towards MORB indicate that the lavas are variably depleted approaching those of MORB-like composition of the oceanic Siletzia LIP. The recognition of the involvement of a depleted source in these compositions is a well-accepted view that has been mentioned in papers for decades [11,16,19,20,25]. Another well appreciated point, typically made for PGB, is that the deviation from a flat incompatible element pattern with enrichment in select LIL elements and LREE is due to elemental overprint of this relatively depleted source [11,16,19,20,60]. The comparison of PGB, AB1&2 north, south, and Steens with typical calc-alkaline mafic lavas of the young Cascade volcanic arc supports this view, and now, not only for PGB but also for the other earliest lavas. The Cascade volcanic arc at 17 Ma was located near where the arc is now (Figure 1) [76]. The distance to the arc at 17 Ma is too far for this overprint to be contemporaneous with CRB magma production, and hence it must have been imparted by earlier subduction-related processes. This in turn allows us to surmise that these lavas were sourced to large degree shallowly (i.e., <90 km) within the zone where slab-derived fluids (and possibly melts) were able to cause this earlier metasomatic overprint [54,77]. Furthermore, comparison of lavas of this study with Siletzia samples indicates that isotopic composition as well as incompatible elemental patterns do not reveal a clear plume signature, contrary to the high-Ti Siletzia samples. This does not preclude some plume component, but these compositional aspects just do not reveal this. The continental flood basalts of the CRBG mostly erupted through young lithosphere of accreted terranes west of the North American craton [78]. This is relevant as this precludes an overestimation of this upper metasomatized mantle component as modeled for the Karoo flood basalts [79]. In the case of the Karoo basalts, ref. [79] could show that a highly enriched component derived from an old subcontinental thick cratonic lithospheric mantle has a significant compositional leverage to impart a subcontinental lithospheric signature by small degree mixing.

6. Conclusions

Our study focused on reevaluating the compositional relationships of the lower-most flows of the American Bar subunit, AB1&2 units, of the Imnaha Basalt to the Picture Gorge Basalt (PGB). Samples of this study were collected from sites distributed along a north—south distance of ~250 km covering most of the outcrop area of the Imnaha Basalt, including some sites in the north that were originally used to establish the flow stratigraphy of Imnaha Basalt in the 1980s. We added our new data to published data of the Imnaha Basalt and grouped samples into three flow packages: (i) lower American Bar flows (AB1&2), (ii) middle to upper American Bar flows (AB3+), and (iii) flows of the Rock Creek chemical type of the Imnaha Basalt (RC). We used these groups for comparison to PGB lavas, and for evaluating AB1&2 flows internally, to other Imnaha flows and relative to other earliest flows of the CRBG across the province. Our findings are the following:

• AB1&2 have more major and trace elemental compositional similarities to PGB than to overlying AB3+ and RC flows. Because of their unique composition, AB1&2 flows among the Imnaha Basalt stand out and hence should be highlighted as their own Imnaha chemical type;
• AB1&2 samples of the northern sampling area have subtle but distinct compositional differences to AB1&2 samples of the southern area;
• Compositional provinciality among PGB, AB1&2 south, north, and Steens Basalt is expressed in (e.g.) Zr/Y, La/Yb, Ba/Nb, Tb/Yb, and radiogenic isotopes. Observed variations cannot be explained by simple mixtures, but rather source material and conditions are unique to give rise to earliest CRBG lava as recorded by AB1&2 south, north, PGB, and Steens Basalt, yet all record tapping a variably depleted and metasomatized mantle;
• The provinciality of the earliest CRBG lavas across the source area of the province in eastern and northeastern Oregon (i.e., the area delimited by the main dike swarms, cf. Figure 1) suggests that lavas were tapped relatively locally, and emplacement distribution was not widespread enough to disrupt the observed regional variations at this eruption stage;
• Absolute ages of all earliest lavas of the CRBG indicate ~synchronous onset of eruptions: >16.7 Ma for AB1&2 north, 17.1 Ma for AB1&2 south, 17.2 Ma for PGB, and 17.0 Ma for Steens Basalt.
• There is little compositional evidence for a plume component in the first lavas erupting across the CRBG province.